Journal of Chongqing University (English Edition) [ISSN 1671-8224]
Vol. 7 No.1
March 2008
73
Article ID: 1671-8224(2008)01-0073-06
To cite this article: TIAN Feng-chun, JI Yan-li, HAN Liang, KADRI Chaibou. Entropy of images after wavelet transform [J]. J Chongqing Univ: Eng Ed [ISSN 1671-8224],
2008, 7(1): 73-78.
Entropy of images after wavelet transform
∗
TIAN Feng-chun
†
, JI Yan-li, HAN Liang, KADRI Chaibou
College of Communication Engineering, Chongqing University, Chongqing 400030, P.R. China
Received 26 March 2007; received in revised form 17 June 2007
Abstract: We studied the variation of image entropy before and after wavelet decomposition, the optimal number of wavelet
decomposition layers, and the effect of wavelet bases and image frequency components on entropy. Numerous experiments were
done on typical images to calculate (using Matlab) the entropy before and after wavelet transform. It was verified that, to obtain
minimal entropy, a three-layer decomposition should be adopted rather than higher orders. The result achieved by using
biorthogonal wavelet decomposition is better than that of the orthogonal wavelet decomposition. The results are not directly
proportional to the vanishing moment, however.
Keywords: image processing; entropy; wavelet transform; data compression; wavelet bases
CLC number: TP391.4 Document code: A
1 Introduction
7
In information theory, for a discrete source
, the
self-information of symbol
i
which occurs with
probability
()
i
px is defined as
()
)
log
ii
xpx=− .
The method is used to measure information. The
average information per source output, denoted
by
()
X , is
() ()
()
() ()
log
ii
i
XEIX px px==
∑
. (1)
This quantity is called the uncertainty or entropy of the
source. It is related to the distribution of source
probability. Indeed, using this theory to calculate image
first-order estimate of entropy provides insight into
image compressibility. The first-order estimate of
entropy is a lower bound on the compression that can
easily be achieved through variable-length coding. This
†
TIAN Feng-chun (田逢春): PhD; Prof.; Research interests: signal
processing, wavelet theory and application, image processing and
optical information processing; E-mail: fengchuntian@cqu.edu.cn.
∗
Funded by the Natural Science Foundation of China (No.
60472037).
definition is also valid even after wavelet
decomposition in image compression.
Entropy has been used in image restoration. Zero-
order entropy has been applied as a basic scheme to the
bit allocation among the eigen images after principal
component analysis (PCA) transformation in lossy
compression of spectral images. Entropy is also
estimated based on a doubly stochastic generalized
Gaussian model for rate constraints in image
compression, which is also combined with mixture
particle filters for multiple-object tracking [1-2].
Mariaa, et al. [3] combined wavelet and entropy
analysis to evaluate the evolution of diffusion and
correlation effects through scale. They calculated the
entropy of smooth and detail coefficient sets generated
by wavelet transformation of some sample images in
each scale, and obtained measures that allowed them to
evaluate these behaviors. They found that the varying
trend of entropy depends on whether diffusion or
correlation is predominant. When correlation
predominates, entropy decreases, whereas when
diffusion predominates, entropy increases. The effects
of diffusion and correlation are different, however, on
different images. This result can also be seen in our
experiments.
Open access to full text of all articles at http://qks.cqu.edu.cn/